Can the supply of physicians, surgeons, or physical therapists meet future demand for musculoskeletal (MSK) care? Probably not.
To make matters worse, the entire MSK system, which relies on about a quarter million highly trained, artisanal, craft-based caregivers, is not easily scalable.
Artificial intelligence (AI) and machine learning (ML) tools, however, could represent a possible solution.
Mid-Course Assessment on the Road to an AI Future
Two researchers, one from Strava Corporation in San Francisco, Phani Teja Nallamothu, Senior Cloud Engineer II and one from the University of the Cumberlands, Jasmin Praful Bharadiya, Ph.D., former research assistant and current DevOps engineer, published a mid-course assessment of these intelligent tools in orthopedics and spine.
Their study, “Artificial Intelligence in Orthopedics: A Concise Review” appears in the April 2023 edition of the Asian Journal of Orthopaedic Research.
By specialty, here was what they found:
- Joint Reconstruction: AI is being employed, on a testing basis, for, “Imaging analysis for automated diagnosis, implant appraisal and clinical outcomes prediction.” Other applications, which apply across all specialties, but are found in joint reconstruction include, say the study authors, “improving pre-operative workflow for patient-specific implants and implant R&D.” Specific areas of AI testing and development are:
- Automated OA Diagnosis
- OA grading
- Prediction risk of OA progression
- Identification of unstable implants
- Automated implant identification
- Preoperative prediction of hospital length of stay, cost and long-term patient reported outcomes
- Preoperative planning assistance
- R&D/Implant optimization
- Spine: By study volume, spine is one of the most active areas for testing AI and ML tools. According to the study authors, “Predicting postoperative complications and using imaging to diagnose spinal disorders are two of the most investigated areas.” Specific areas of AI testing and development are:
- Automated vertebral localization
- Automated pathology diagnosis
- Disc degeneration severity classified
- Prediction of postoperative complications
- Orthopedic Oncology: “Both primary bone and soft tissue malignancies, as well as metastatic illnesses, have been the focus of AI research.” The researchers noted that the use of AI in orthopedic oncology is still in its infancy, but that the outcomes reported in the literature are promising. Specific areas of AI testing and development are:
- Automated diagnosis of tumors
- Classifying tumors as benign or malignant
- Prediction of recurrence, survival, and life-expectancy
- Tumor burden analysis
- Trauma: For trauma applications, AI image technologies are gaining the most traction. Specifically, “Automated image-based fracture diagnosis is where orthopedic trauma applications now stand.” As in other specialties, AI in trauma care is still in its early stages but “research on the use of AI to predict clinical outcomes for victims of trauma is beginning to emerge.” Specific areas of AI testing and development are:
- Automated fracture detection
- Sports Medicine: The study authors found that “Anterior cruciate ligament (ACL) and meniscal tear identification is the most common application in the study of knee injuries.” Since sports medicine is, by and large, a soft tissue repair practice, imaging—notably MRI imaging—is especially critical to accurate diagnosis—which, of course, lends itself well to AI tools. Specific areas of AI testing and development are:
- Automated detection of soft tissue pathology
Application of AI by Anatomical Region
The authors, as part of their literature review, tracked which anatomic region was most often the target of AI tools. Their conclusions were summarized in the following chart:
Here, from the study is a partial list of the musculoskeletal AI clinical studies referenced in their paper:
- Automatic identification of adolescent idiopathic scoliosis through optimization of a three-dimensional spine model vector. Thong W. et al. Three-dimensional morphology study of surgical adolescent idiopathic scoliosis patient from encoded geometric models. Eur Spine J. 2016;25 (10):3104-3113.
- Radiographic image analysis for fracture diagnosis. Olczak J, et al. Artificial intelligence for analyzing orthopedic trauma radiographs. Acta Orthop. 2017;88(6):581-586.
- In order to make accurate clinical forecasts in the future, ML-based predictions for physician order input have shown that it is preferable to priorities smaller amounts of more recent data over bigger amounts of older data. Chen JH, et al. Decaying relevance of clinical data towards future decisions in data-driven inpatient clinical order sets. Int J Med Inform. 2017;102:71-79
- Dual-energy x-ray absorptiometry for the prediction of hip fractures. Kruse C, Eiken P, Vestergaard P, Machine learning principles can improve hip fracture prediction. Calcif Tissue Int. 2017;100(4): 348-360.
- The mechanical performance of a short-stem total hip replacement ban be improved by using machine learning techniques. Cilla M, et al. Machine learning techniques for the optimization of joint replacements: Application to a short-stem hip implant. Plos One. 2017;12(9):e0183755.
- New York University uses a value-based care-aligned artificial intelligence system (PersonaCARE) to manage its middle-aged and elderly fracture population. Karnuta JM, et al. Bundled care for hip fractures: A machine-learning approach to an untenable patient-specific payment model. J Orthop Trauma. 2019;33(7): 324-330.
- Having concluded that the current value-based bundled care approach to hip fractures is unsustainable. Ramkumar PN, et al. Preoperative prediction of value metrics and a patientspecific payment model for primary total hip arthroplasty: Development and validation of a deep learning model. J Arthroplasty. 2019;34(10):2228-2234.
- Joint replacement of the lower extremities: expected hospital expenditures, duration of stay, and patient outcome. Shah RF, et al. Variation in the thickness of knee cartilage. The use of a novel machine learning algorithm for cartilage segmentation of magnetic resonance images. J Arthroplasty. 2019;34(10): 2210-2215.
- Articular cartilage thickness in MRI of healthy knees, automatically measured and segmented. Harris AHS, et al. Can machine learning methods produce accurate and easy-to use prediction models of 30-day complications and mortality after knee or hip arthroplasty? Clin Orthop Relat Res. 2019;477(2):452-460
- Estimation of Complications and Mortality Rates 30 Days After TJA. Hamlet WP, Fletcher A, Meals RA. Publication patterns of papers presented at the annual meeting of the American Academy of Orthopaedic Surgeons. JBJS. 1997;79(8):1138-43
- Predicting in-house SNF utilization after TJA with ANN using internal EMR data. Fontana MA, et al. Can machine learning algorithms predict which patients will achieve minimally clinically important differences from total joint arthroplasty? Clin Orthop Relat Res. 2019;477(6):1267- 1279 and Thirukumaran CP, et al. Natural language processing for the identification of surgical site infections in orthopaedics. J Bone Joint Surg Am, 2019. 101(24): p. 2167- 2174
- Using ML before TJA surgery to determine which patients have a good chance of experiencing MCI is a promising area of research. Thirukumaran CP, et al. Natural language processing for the identification of surgical site infections in orthopaedics. J Bone Joint Surg Am, 2019. 101(24): p. 2167- 2174
- Infections at orthopedic surgery sites can be detected using natural language processing. Galbusera F, Casaroli G, Bassani T. Artificial intelligence and machine learning in spine research. JOR Spine. 2019;2(1): e1044.
- Excellent summary of the use of AI and ML in spine studies. Myers TG, et al. Artificial intelligence and orthopaedics: An introduction for clinicians. J Bone Joint Surg Am. 2020;102(9):830- 840
Conclusions
The field of AI and ML in orthopedics is entering its more creative phase, I think, and the range of ideas, particularly in trauma, is very encouraging. Noteworthy, as well, is that the authors, while hailing from academia, are now in industry, in Silicon Valley. From that frame of reference, it is refreshing to see how they see the contours of an AI and ML musculoskeletal future.
